Re: /var mounted noexec
Nathan E Norman writes: > On Fri, 10 Sep 1999, Matus "fantomas" Uhlar wrote: >> can anyone tell me where does dpkg store its preinstall and >> postinstall scripts when installing packages ? > > /var/lib/dpkg/info iirc That is correct. However dpkg is not necessarily the only program which executes scripts from within /var. atd, for example does the equivalent of ``/bin/sh
Re: #!/Perl question
Brad writes: > One solution would be to write a makefile that detects the location > of the perl binary and then prepends the proper shebang line to the > beginning of your perl programs. Another trick is this: #! /usr/bin/env perl print "Hello, world\n"; This depends on env being in /usr/bin everywhere, but perhaps that's more widely true than it is of Perl. Personally I just use `#! /usr/bin/perl'. ttfn/rjk
Re: [C|E]Tags recursively
Paulo Henrique Baptista de Oliveira writes: > Hi all, > I want to transverse /usr/src/linux with Ctags or Etags and there isnt >an option to do this. > My question is: how to do this thing? Something along these lines: find . -name "*.[ch]" -print | etags - should do the trick. ttfn/rjk
Re: More on XDM and runlevels
Chad Walstrom writes: > Well, I can't site any reason w/xdm other than possibly using > something like kdm or gdm periodically. However, I do know run > levels come in handy. Let's say, for example, you'd like to work on > your web pages at home, perhaps running mysql and apache with php. > Yet, you're only running a 486 which has limited resources. You've > got no need to run Apache and MySQL all of the time, just when you > need them. There's where run levels come into play. Another example is to reduce the level of service during backups, when you don't want any activity that will change files half way through them being copied to tape, etc. Whether such activity is likely can depend on your exact situation. ttfn/rjk
Size of Debian
I unpacked all the *.orig.tar.gz from .../bo/source, and did a little counting. du reports 1.7Gb in 773 packages. I did some line counts as well; we have more than 52 million lines in total, of which over 18 million are in *.c files and over 3 million in *.h files. -- Richard Kettlewell http://www.elmail.co.uk/~richard/ -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED] .
X and modmaps
I'm attempting to exchange the meanings of my left CTRL key and my Caps Lock key, as I have recently got used to an X terminal where they are in different physical locations to the PC keyboard. After determining the keycodes generated by these keys, I set them with xmodmap as follows: keycode 66 = Control_L keycode 37 = Caps_Lock xev suggests this has succeed: KeyPress event, serial 18, synthetic NO, window 0x401, root 0x2a, subw 0x0, time 1391234933, (57,141), root:(877,163), state 0x0, keycode 66 (keysym 0xffe3, Control_L), same_screen YES, XLookupString gives 0 characters: "" KeyPress event, serial 21, synthetic NO, window 0x401, root 0x2a, subw 0x0, time 1391235358, (57,141), root:(877,163), state 0x2, keycode 37 (keysym 0xffe5, Caps_Lock), same_screen YES, XLookupString gives 0 characters: "" However, attempting to use them e.g. in Emacs or an xterm, they still have their unexchanged meanings. What am I doing wrong? -- Richard Kettlewell http://www.elmail.co.uk/~richard/ Would it be so terrible, to be a photino bird? -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED] .
Re: Return code from system()
"Oliver Elphick" writes: >According to the man page for system(), it should return the return >code of the command run, or else -1 or 127. > >For me it is returning RC * 256: that is, if the return code is 1, >system() returns 256; if 2, 512 and so on. > >Code fragment: > > char *s = "..."; > int j = system((const char *) s); > >dpkg -s libc5: > >Version: 5.4.20-1 > >Is this a fault in documentation, or in the libc package? Is it unique to me? It's returning the value it gets from wait(2), the man page for which documents some macros for decomposing it. It's a bug in the documentation. -- Richard Kettlewell http://www.elmail.co.uk/~richard/ Are you sure you want to quit reading news? (y or n) -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED] .
Re: zsh vs bash
Both are excellent interactive shells. >* automatic completion on variables names, e.g. type > "export DISP" and hit tab. (I just checked, in bash you can use > Esc-$ to specifically complete a variable name; in zsh the default > compctl (completion) has been setup to complete for a variable name > if the command is "export". While the zsh seemed easier, I guess > the bash approach allows you to control it more.) IIRC bash will complete over variable names when you write a $ at the start. zsh has other cute completion features, e.g. I have this in my .zshrc: hosts=(valour cushioned myrddin tacitus chiark \ mercury.elmailer.net wigwam.elmail.co.uk sunsite.doc.ic.ac.uk \ ftp.uu.net ftp.sendmail.org tlingit.elmail.co.uk) ssh=(chiark) compctl -k ssh ssh compctl -k hosts telnet ftp rlogin rsh ping traceroute compctl -k hosts -f rcp scp which gives me hostname completion on a selection of network commands. >There are 2 programs that really pay off putting a bit of effort into >learning: the shell you use and the editor you use. Picking a "simple >and easy to use" editor is a short sighted approach. Pick a powerful >editor and invest some time in learning it. (You don't have to learn >it all, and you don't have to learn all that much at first, either.) >It will really pack off. And I think the same philosophy applies, >perhaps to a lesser degree, to the shell you use. Agreed. -- Richard Kettlewell http://www.elmail.co.uk/~richard/
Re: Mail and News with UUCP over TCP/IP
>Question for Richard: do you have to have a static IP address assigned to >you to use UUCP over TCP/IP? No, you don't - only the server does. When uucico runs on the client machine, it makes a TCP/IP connection to port 540 of the server machine (which would normally be a machine permanently connected to the Internet). >From then on in, the sequence of events is the same as if it had connected by dialing over a modem without an intervening IP stack: the client machine logs into the server in the traditional UNIX way and speaks a UUCP protocol over the connection. There is a potential oddity, depending how things are set up. With the UUCP services run by my employer, if uucico goes in over TCP/IP, the chat script uses only one login/password pair; whereas if it goes in over dialup, there's two - one to authenticate to the Ascend which actually picks up the phone, and one to authenticate to the UUCP server machine. The Ascend just makes a port 540 connection to the UUCP machine, so while it looks like UUCP over dialup to the customer, it appears to be UUCP over TCP/IP to the server. Not that the software cares much either way. (This is also true of PSI UK's service, in fact.) -- Richard Kettlewell http://www.elmail.co.uk/~richard/ [wubba wubba wubba wubba wubba wubba wubba wubba] It was a creepy and surreal morning when they implanted the biochips in the mind of Mohinder Singh. [wubba wubba wubba wubba wubba wubba wubba wubba]
Re: Mail and News with UUCP over TCP/IP
Stefan Walder writes: >I want to use my Linux-Box at home to get mail and news. I'm using my >Uni-ISP and i want to have several Email-Adresses (my family). And I >don't want to use POP, I think uucp is nicer! So I want to use uucp >over TCP/IP! Now my questions: Particularly for multiple addresses it has its advantages, yes. I've some experience of setting it up under SVR4... It works, though `nice' is not a word I'd use l-) > 1.) Can I use the email-address [EMAIL PROTECTED] That's really a question for your ISP; they will need to take some action to route mail for whatever domain they give you into your UUCP queue, you can't normally expect to just have things work by choosing the right address format. > 2.) What does I need to get and post news? I think I need suck, but is this > sufficient? Depends ... there are a variety of programs with names like `slurp' and `suck' which retrieve news from an NNRP server and (e.g.) inject it into your own news server. They're quite easy to write. You'll need NNTP access to feed news that you post back. I'm assume you'll be running a local copy of INN, though this is not the only way to do it. If you want to do the *whole* thing over UUCP that's possible too, your ISP will set up an outgoing UUCP feed and you must do the same to feed news back. Last time I tried to do this it was too hard so I gave up and just fed news back out by NNTP instead - a good thing I've never had to do it without TCP/IP being available... -- Richard Kettlewell http://www.elmail.co.uk/~richard/ [wubba wubba wubba wubba wubba wubba wubba wubba] It was a creepy and surreal morning when they implanted the biochips in the mind of Mohinder Singh. [wubba wubba wubba wubba wubba wubba wubba wubba] -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
Re: problem with suid perl scripts in Debian 1.2
Christopher E. Stefan writes: >After upgrading from 1.1 to 1.2 none of my suid-perl scripts worked >anymore. I ended up having to put C wrappers around all of them. I reported this as a bug a while back, don't recall the number. A bit disappointing to see so little reaction. -- Richard Kettlewell http://www.elmail.co.uk/staff/richard/ -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
Re: innd & mmap
>> my news box is regularly throttling with the log message "File exists >> writing symlinking article file -- throttling". Try rebuilding your history file - IIRC this has worked for me (albeit under Solaris 2.5 rather than Linux). >> Is this still true for linux? >> >> is it a kernel or a libc problem? >> >> Has inn for debian been compiled with MMAP turned off? If not, should >> this be reported as a bug against inn? > >Hmm.. I run a non-debian version of inn, aqnd it works fine with mmap. Ditto. -- Richard Kettlewell http://www.elmail.co.uk/staff/richard/ /dev/hdb5 - 0.5Gb of spinning metal, all alone in the night... -- TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to [EMAIL PROTECTED] . Trouble? e-mail to [EMAIL PROTECTED]
Re: seg faults with majordomo
In article <[EMAIL PROTECTED]> you write:
>
>help help :/
> i've been tring to setup majordomo and have done the following:
>
>1 made '/etc/smail/lists' dir
>2 did ln -s /var/lib/majordomo/lists /etc/smail/lists
>3 added to /etc/smail/directors:
> aliasinclude:
> driver=aliasinclude,nobody;
> copysecure,copyowners
>and
> lists:
> driver=forwardfile,
> caution,
> nobody,
> sender_okay,
> owner=owner-$user;
> file=lists/${lc:user}
>
>4 made sure the proper aliases were in the /etc/aliases file
> majordomo: "|/usr/lib/majordomo/wrapper majordomo"
> list-request: "|/usr/lib/majordomo/wrapper majordomo"
> majordomo-owner: adren
> owner-majordomo: adren
>and
> hottub-owner: adren
> hottub-approval: adren
> owner-hottub: hottub-owner
> hottub-request: majordomo
> hottub: :include: /etc/smail/lists/hottub
^
I think you don't want a space there. You may need to quote the whole
address, as well.
>5 made a list-file 'hottub' in /var/lib/majordomo/lists which is also in
>the /etc/smail/lists dir
Have you read the majordomo `README' file wrt permissions, etc. Note
that recent versions of the `base' package broke the Majordomo
package, don't know if it's been fixed yet.
> when sending mail to majordomo for help or lists it sends the
>correct info back to me. if i send to the list i get the mail returned
>with 'no such recipient'.
Might need to see the full bounce.
Things work fine for me without messing around with /etc/smail/lists
or the lists director.
--
Richard Kettlewell [EMAIL PROTECTED] [EMAIL PROTECTED]
http://www.elmail.co.uk/staff/richard/
Re: MajorDomo Problem with "wrapper"
Brian C. White writes:
>I'm trying to set up majordomo on our server, but have run into the
>following problem.
>
>Any mail sent to the "majordomo" alias and gets piped into the majordomo
>command via "wrapper" causes majordomo to just spin its wheels, chewing
>cpu cycles and allocating more and more memory. If I run the command
>manually ("wrapper majordomo off the wrapper and do (as root) "majordomo
>My alias:
>
>majordomo: "|/usr/lib/majordomo/wrapper majordomo"
>
>
>Any ideas?
Check the ownerships and permissions on the log directories. The log
directory must be 775 majordom.majordom, and the logfiles in it 664
majordom.majordom.
- Richard
Re: aout svgalib problem
David Engel writes: >Richard Kettlewell writes: >>Jeffery S. Coy, Jr. writes: >>>i just installed the aout-svgalib-1.28-6.deb package, and noticed it >>>installs to /usr/i486-linuxaout rather than /usr/lib/i486-linuxaout, so >>>the system can't find it. >> >>No, /usr/i486-linuxaout/lib is the correct place; I'd expect the >>problem to be something else. I don't really have time to maintain >>the svgalib packages any more and it would be good if someone else >>would take over. > >This is partly my fault. In the last libc4 package, I moved libdb.so* >from /usr/i486-linuxaout/lib to /usr/lib/i486-linuxaout. I did this >so users who didn't install any a.out development packages >(eg. libc4-dev, aout-gcc, etc.) wouldn't have a /usr/i486-linuxaout >directory cluttering up /usr. Oh. I wasn't aware of the change, though I think it a good one. If it was announced anywhere then I missed it. It would have been a good idea to tell the runtime linker to continue searching the old place, though. Doing this should be an adequate workaround for the time being, I believe. I think I've quite a bit of catching up to do on change file formats, package uploads and so on, so (if no-one else does the job) it could be a while before an updated version appears. - Richard -- http://www.elmail.co.uk/staff/richard/ GCS d- s+:- a-- C++ ULVS+++$ P+++ L++ E++ W(++,--) N(++,+) o? K w--- O? M- V? PS(+,+++) PE Y+ PGP+ t- 5++ X+@ R tv--- b++> DI+ D+ G e++ h r% y++
Re: aout svgalib problem
Jeffery S. Coy, Jr. writes: >i just installed the aout-svgalib-1.28-6.deb package, and noticed it >installs to /usr/i486-linuxaout rather than /usr/lib/i486-linuxaout, so >the system can't find it. No, /usr/i486-linuxaout/lib is the correct place; I'd expect the problem to be something else. I don't really have time to maintain the svgalib packages any more and it would be good if someone else would take over. >i tried writing the maintainer ([EMAIL PROTECTED]), but the host >appears not to exist. Circumstances beyond my control and all that, I'm afraid. I hadn't realized that that address had made its way into package maintainer fields. We do have a bug tracking system for a good reason, BTW. - Richard -- http://www.elmail.co.uk/staff/richard/ GCS d- s+:- a-- C++ ULVS+++$ P+++ L++ E++ W(++,--) N(++,+) o? K w--- O? M- V? PS(+,+++) PE Y+ PGP+ t- 5++ X+@ R tv--- b++> DI+ D+ G e++ h r% y++
Re: pppd and setuidness (was Re: 3 Questions)
>>One could make the uid of the account zero to achieve this without >>making pppd setuid, though I can imagine this making people jump up >>and down about security - can anyone think of an attack on this? > >If the user figures a way to change their shell, you're dead. Quite so. Similarly if there's a way of running a shell under a uid provided you know the appropriate password. AFAICT su is safe against this as long as pppd (or whatever script one uses) isn't in /etc/shells. I'd be very wary indeed of actually *trying* this! - Richard -- http://www.elmail.co.uk/staff/richard/ GCS d- s+:- a-- C++ ULVS+++$ P+++ L++ E++ W(++,--) N(++,+) o? K w--- O? M- V? PS(+,+++) PE Y+ PGP+ t- 5++ X+@ R tv--- b++> DI+ D+ G e++ h r% y++
Re: Majordomo install problem
Juhani Luhtanen writes: >Failed to create group majordom: >adduser: the user you specified already exist > >when I try to run dpkg --install with the majordomo.deb file. It looks to me like the base disks now have the `majordom' user but not the corresponding group (before you even install Majordomo, that is). Can someone confirm or refute this please - I don't have time to check right now. Since the base disks for 0.93 had neither the user nor the group the preinst attemps to create them. It then checks that both are there. If that is the problem ... as a temporary workaround create a `majordom' group, preferrably with the same GID as the `majordom' group. Make the default group for the `majordom' user be the GID of the `majordom' group. Then try reinstalling the majordomo package. - Richard -- http://www.elmail.co.uk/staff/richard/ GCS d- s+:- a-- C++ ULVS+++$ P+++ L++ E++ W(++,--) N(++,+) o? K w--- O? M- V? PS(+,+++) PE Y+ PGP+ t- 5++ X+@ R tv--- b++> DI+ D+ G e++ h r% y++
pppd and setuidness (was Re: 3 Questions)
>>good question. and why isn't pppd setuid root? if it's a security issue, >>a ppp group would be in order. > >I'd say 'because it doesn't neeed to be' is a good justification. > >If you need to have non-root users execute ppp as root, take a look >at the 'sudo' or 'super' packages. They allow you to define commands >that can be executed as root by a set of users... without forcing >your choice of 'this should be setuid root' programs on all other >Debian users. Personally I find that the diald package is an excellent way of avoiding this whole issue. However: As someone pointed out last time this came up, when your machine is acting as a PPP server you need to run the pppd as root from a dialin account. One way of doing this (with the commercial PPP with which I am familiar) is to make the pppd setuid root and run it from a shell script which is that user's login shell. (I suppose you could run the pppd directly, but doing it from scripts is more convenient as it allows you to pass arguments to the pppd and set various options on a per-user basis.) One could make the uid of the account zero to achieve this without making pppd setuid, though I can imagine this making people jump up and down about security - can anyone think of an attack on this? - Richard -- http://www.elmail.co.uk/staff/richard/ GCS d- s+:- a-- C++ ULVS+++$ P+++ L++ E++ W(++,--) N(++,+) o? K w--- O? M- V? PS(+,+++) PE Y+ PGP+ t- 5++ X+@ R tv--- b++> DI+ D+ G e++ h r% y++
Re: Interesting Activity for Shell commands.
Martin Rheumer writes: >I am running a number of shell scripts and find I have to specify sh >and then the script name. The first line of the script is always >#!/bin/sh which I am assuming takes care of this. Could someone >provide some hints what I have done wrong or some pointers. Have you made the script executable? chmod +x scriptname is perhaps what you want. You may also have trouble if the script isn't in your PATH; try ./scriptname assuming the script is in the current working directory. - Richard -- http://www.elmail.co.uk/staff/richard/ GCS d- s+:- a-- C++ ULVS+++$ P+++ L++ E++ W(++,--) N(++,+) o? K w--- O? M- V? PS(+,+++) PE Y+ PGP+ t- 5++ X+@ R tv--- b++> DI+ D+ G e++ h r% y++
Re: Cron
>How can I stop cron sending a message root everytime it runs a >script. I have a couple of cron jobs that run every 15 minutes and I >am getting hundreds of messages a day. Make sure that the programs in question don't produce any output (e.g. by redirecting the output to /dev/null.) Cron mailing you any output that it sees is a feature. - Richard -- http://www.elmail.co.uk/staff/richard/ GCS d- s+:- a-- C++ ULVS+++$ P+++ L++ E++ W(++,--) N(++,+) o? K w--- O? M- V? PS(+,+++) PE Y+ PGP+ t- 5++ X+@ R tv--- b++> DI+ D+ G e++ h r% y++
Re: perl and setgid scripts
...in fact apropos my previous comments about perl, I note that the description reads as follows: Larry Wall's Practical Extracting and Report Language. An interpreted scripting language, known among some as "Unix's Swiss Army Chainsaw". Perl is optimized for scanning arbitrary text files and system administration. It has built-in extended regular expression matching and replacement, a dataflow mechanism to improve security with setuid scripts and is extendible via modules that can interface to C libraries. It's not clear what point there is in advertising the taint feature if you're not going to implement set[ug]id scripts... Darren? - Richard -- http://www.elmail.co.uk/staff/richard/ GCS d- s+:- a-- C++ ULVS+++$ P+++ L++ E++ W(++,--) N(++,+) o? K w--- O? M- V? PS(+,+++) PE Y+ PGP+ t- 5++ X+@ R tv--- b++> DI+ D+ G e++ h r% y++
perl and setgid scripts
I used to be able to do setgid (and indeed, setuid) scripts in perl; but after upgrading to 5.002-7, they break, keeping the uid/gid of the invoking process. This has just spannered my mail system, which uses a perl script to do the final delivery (and delivery agents must be setgid mail under Debian). What gives? - Richard -- http://www.elmail.co.uk/staff/richard/ GCS d- s+:- a-- C++ ULVS+++$ P+++ L++ E++ W(++,--) N(++,+) o? K w--- O? M- V? PS(+,+++) PE Y+ PGP+ t- 5++ X+@ R tv--- b++> DI+ D+ G e++ h r% y++
Re: Must pppd be run by root?
I think `net' would be a good group name for this. Rob Browning writes: >> "C" == Craig Sanders <[EMAIL PROTECTED]> writes: > >C> I suppose that would do the job, but what if a sysadmin wants to >C> allow users to dial in using ppp, but NOT allow them to dialout >C> with minicom or send faxes? > >C> I'm absolutely certain that I wouldn't want to add users to a >C> dialout group just to let them dial in with a ppp account. > >If you're going to do something, ppp/slip seem sufficiently different >to warrant their own group. Something like >transient-network-connection (but shorter :> ). > >I'm still not sure that ppp will really work properly when run by >someone other than root (unless it's suid root) if you want to use the >"default" option that makes the routing changes automatically when the >connection goes up and down. Even if you handle the routing outside >the call to pppd, you'll still have to do the routing as root, so I >don't see what trying to get pppd to be runnable by non-root users >buys you. > >I still think that sudo's the best solution, but I probably just don't >understand the situation well enough. > >-- >Rob > >
Re: find question (and xargs)
>>find / -size +459976c -noleaf -type f -name '*.deb'|\
>>xargs -n 1 dpkg-split -s {} && rm {}
>>
>>I was thinking that {} would be replaced by the filename but that's
>>not the case. Anyone know how to solve this?
>
>It's find that does the replacing. None of the {}s are in the find
>arguments, however. (And rm is not even in the xargs arguments!)
Using `xargs -i' will substitute for {}.
To get && right, put "quotes" round it.
>Personally, I'd probably make a script for the split-and-remove, but it
>should also work with a shell function.
Doesn't work for me, nor did I expect it to.
ttfn/rjk
Re: Must pppd be run by root?
Craig Sanders writes: > >On Sun, 12 May 1996, Rob Browning wrote: > >> >>>>> "R" == Richard Kettlewell <[EMAIL PROTECTED]> writes: >> >> R> pppd has to do various messing around creating network interfaces >> R> and so on, so running it as (not root) is a bit of a non-starter. >> R> Why do you want it to be able to run it not as root? > >two words: > >DIAL-IN USERS > >> Right, pppd really needs to be run as root. You don't want just any >> user killing your net connection. If you want to let a selected set >> of users bring the net connection up and down, check out sudo. It's >> designed for this. > >This permissions problem just cost me over $44 (australian dollars). > >I upgraded pppd on the dialin machine at work on Friday night. Forgot to >change the permissions on pppd to make it owned by root.ppp, perms=1750. > >When the line dropped out from my home machine to my work machine at 5am >the next morning, my home system automatically dialed in Under the control of what user? Why not root? >to re-establish the link. Logged in OK, ran pppd. The connection >dropped again because it didn't have the right perms. My system then >proceeded to call every 2 minutes, until I got home and stopped it at >about 11am. ttfn/rjk
Re: diald (Was: Re: Must pppd be run by root?)
I think this was meant to be sent to the mailing list. --- start of forwarded message (RFC 934 encapsulation) --- From: Yves Arrouye <[EMAIL PROTECTED]> To: [EMAIL PROTECTED] (Richard Kettlewell) Subject: Re: diald (Was: Re: Must pppd be run by root?) Date: Mon, 13 May 1996 00:17:42 +0200 Richard Kettlewell writes: > The diald package removes all the hassle of starting and shutting down > network connections over a transient link, incidentally - you might > like to investuigate that. Is there a debianized version of diald 0.14 out there? (ELF). With the dctrl tk tool, too! (I can email it to the one making the package if needed). Also, do someone know how to ask sendmail really not to make DNS queries? This is somewhat that starts up diald when one does not want, and it is really annoying. A last question for you diald gurus: I used diald and after having asked it to close the connection (using its control fifo) I still can see monitored packets associated with a time to close the link. Anybody knows why? I don't use ftp, http or anything after having asked to close... Yves. --- end ---
Re: Must pppd be run by root?
Rick Macdonald writes: >I'm configuring a fresh install of 1.1. I can only start pppd as >root. I found that /etc/ppp had permisions of 700. After changing >that it still can only be run as root. Comparing "strace" output >shows that it may be failing doing an ioctl on /dev/ttyp0. Changing >the permissions of this tty to 666 didn't help. > >Any ideas? pppd has to do various messing around creating network interfaces and so on, so running it as (not root) is a bit of a non-starter. Why do you want it to be able to run it not as root? The diald package removes all the hassle of starting and shutting down network connections over a transient link, incidentally - you might like to investuigate that. ttfn/rjk
Re: Probs with 'tail -f foo | awk ...'
>Since upgrading from Slack 2xxx to Debian 1.1 the following >construction doesn't work any more :-( > >adson$ cat > test << EOF >>A 1 >>B 1 >>C 1 >>EOF > >adson$ tail -f test | awk '$1 =="B"' > >There should be the output "B 1" on stdout now, but nothing >happens. It works if I don't use the f-switch, but this is not what I >need. > >Any hints?! I'd guess `tail -f' is buffering its output, and so not writing to the pipe until it has a full block. Or, awk is buffering input and waiting for more to come out of tail. Or both. Try with a much bigger `test' file and see what happens. -- Richard Kettlewell http://www.elmail.co.uk/staff/richard/[EMAIL PROTECTED] If my head is in the clouds, it is because I am standing on the shoulders of giants...
Re: Ftpd annoyance - DIR doesn't work for anonymous ftp
>I've just enabled anonymous ftp to my Debian 1.1 Linux box according >to the instructions in the ftpd manpage. All works fine, except that >it appears that an anonymous ftp user cannot use the command "DIR" >and "ls -l" doesn't work (both report no files). This works fine for >regular users. I simply copied /bin/ls to ~ftp/bin; I thought the >problem may have been that ls couldn't find its libraries after the >chroot, but copying libc.so to ~ftp/lib didn't help either. What about ld.so? You might find some error messages in log files, with any luck: those would provide a pointer, too. -- Richard Kettlewell [EMAIL PROTECTED]http://www.elmail.co.uk/staff/richard/
Re: gs shouldn't depend on svgalib
joost witteveen writes: >>The gs package is compiled to use svgalib and X. This is a nuisance, I >>had to install svgalib only for gs. I propose to split it in one for >>svgalib only and another for X only. Consider also the cases where the user wants svgalib but not X; and also the case where they want both X and svgalib. That's three packages, or six if there're multiple versions of gs... As for it being a nuisance, won't dselect just do the right thing for you automatically? >I'm sorry, but this is the way it has been decided to do it for all >packages that could use more libs: simply make them decide on all >libs. This makes installation much easier, while not adding too much >disk space. Splitting gs in two would create 4 .deb files (2 for >gs-2.62, 2 for gs-3.53). I'd rather answer about 2 emails/half year >about gs depending on both svgalib and X than having to explain what >4 gs packages is good for. I agree with this. -- Richard Kettlewell (Debian svgalib maintainer) http://www.elmail.co.uk/staff/richard/[EMAIL PROTECTED] If my head is in the clouds, it is because I am standing on the shoulders of giants...
